Comments on: Lessfs-0.8.0 is available for download.

By: Chris

Chris — Fri, 06 Nov 2009 07:08:58 +0000

Hi Mark,

unbelievable, your support is already better than the one at big companies

The 0.8.1 works great, also the inplace-update from 0.7.5 worked flawlessly.

Thanks again for your great work!

Chris

By: Mark Ruijter

Mark Ruijter — Thu, 05 Nov 2009 21:27:11 +0000

Hi Chris,

Just include this line somewhere in lessfs.cfg and your problem is gone:
DEBUG=2

This problem is actually a know bug. lessfs-0.8.1 fixes it.
mklessfs will segfault when DEBUG is not set.

Mark.

P.S. I have attached a ‘pre-release’ of lessfs-0.8.1 for your convinience.

By: Chris

Chris — Thu, 05 Nov 2009 21:00:37 +0000

Hi Mark,
Sorry I have to disappoint you but since I had some spare time I just formatted the whole system and did a clean installation. I had played with the old system around a lot to find the best setup, so I had that planned anyway.
However my expectation towards 0.8.0 on a physical machine was wrong, I still get the memory access error when trying to use the mklessfs. I am sure the config has not changed.
As said this error did not occur in a VM, though I was using the exact same sources for everything.
Maybe I can at least help out in this case?
Chris

By: Mark Ruijter

Mark Ruijter — Thu, 05 Nov 2009 20:53:35 +0000

Hi Chris,

Do you still have the corrupted databases?
There are several ways to check them. I have almost completed lessfsck
for the tokyocabinet backend so you could try that.
Did you check /var/log/messages? Did it lessfs log something?

Are you sure that lessfs.cfg did not change during the upgrade?
Can you email me your lessfs.cfg?

I hate loose ends.

Mark.

By: Chris

Chris — Thu, 05 Nov 2009 13:56:42 +0000

Hi Maru,
thanks for the hint, unfortunately it did not work as expected. After I had compiled and installed 0.8.0 (previous version was 0.7.5) the database must have somehow gone corrupted, I had duplicate folders in various directory listings (identical attributes) but was only able to access one copy. Also I wasn’t able to delete anything any more, the “socket connection failed” repeatedly.
Completely removing everything and reinstalling 0.8.0 did not work either. Whenever I tried to run mklessfs with my config file it kept saying “Speicherzugriffsfehler” (German version, translates to something like memory access error).
I am using the latest Debian 64bit. Surprisingly a clean Debian installation inside a VM works just fine and probably will work on a physical machine, too. This is just to let you know there might be some bug, don’t bother too much about that
Chris

By: maru

maru — Wed, 04 Nov 2009 17:07:43 +0000

Hi Chris,

When you use the tokyocabinet data store (not file_io) upgrading to the upcoming releases should work without a problem.
If you have a rpm based distribution rpm -Uvh will do the trick. Most distributions are now going to include lessfs so you can use the updates that come with your distro.
I know that Suse and Fedora have lessfs rpm packages.

I just started to work on fsck for the file_io backend, fsck for tokyocabinet is pretty much done. To enable a good sanity check I may be forced to change the file_io data layout.
This is why the code is still beta I guess.
Once fsck works for file_io and tc lessfs-1.0 stable will be released. I think this will be the case at the end of this month.

Only when people report critical bugs the coming weeks this schedule may be delayed.

By: Chris

Chris — Wed, 04 Nov 2009 15:50:39 +0000

Thank you for this great filesystem, this makes my backups at home so much easier
Anyway I have still got a question (which might arise because I am pretty new to Linux):
Is there a way to move from one version/release to another without losing all the data in the old version and without having to copy the contents to the new version, in short an in-place update?

By: Dimitri Bellini

Dimitri Bellini — Tue, 03 Nov 2009 09:56:00 +0000

Hi Mark
i have made my test and i must to say there is another improvement!
On write a new object there is speed enhancement of 3 seccond compared to old 0.7.4, the only strange behaviour is on write a DIVX file because in the old 0.7.4 a write this file (717410KByte) in 126 sec now with the new 0.8.0 i have 187 sec.
Other good news i have notice other big speed improvement on read a deduped data ex. on 0.7.4 the transfer rate was 11387 (KByte/Sec) now 13796 (KByte/Sec).
Many thanks
For your great work
Dimitri